- Email: [email protected]
CT Data on Esophageal Target Definitions in Radiation Planning
Int. J. Radiation Oncology Biol. Phys., Vol. 72, No. 5, pp. 1612–1618, 2008 Copyright Ó 2008 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/08/$–see front matter
doi:10.1016/j.ijrobp.2008.07.061
PHYSICS CONTRIBUTION
IMPACT OF MANUAL AND AUTOMATED INTERPRETATION OF FUSED PET/CT DATA ON ESOPHAGEAL TARGET DEFINITIONS IN RADIATION PLANNING THEODORE S. HONG, M.D.,* JOSEPH H. KILLORAN, PH.D.,yz MARCELO MAMEDE, M.D., PH.D., y AND HARVEY J. MAMON, M.D., PH.D. * Department of Radiation Oncology, Massachusetts General Hospital, Department of yRadiation Oncology and z Nuclear Medicine, Brigham and Womens’ Hospital, Boston, MA Purpose: We compare CT-only based esophageal tumor definition with two PET/CT based methods: (1) manual contouring and (2) a semiautomated method based on specific thresholds. Methods and Materials: Patients with esophageal cancer treated at Brigham and Women’s Hospital from 2003 to 2006 were identified. CT-based tumor volumes were compared with manual PET/CT-based volumes and semiautomated PET-based tumor volumes. Differences were scored as (1) minor if the superior or inferior extent of the primary tumor (or both) differed by 1–2 cm and (2) major if the difference was > 2 cm or if different noncontiguous nodal regions were identified as being grossly involved. Results: Comparing CT-based gross tumor volumes (GTVs) to manually defined PET/CT-based GTVs, use of PET changed volumes for 21 of 25 (84%) patients: 12 patients (48%) exhibited minor differences, whereas for 9 patients (36%), the differences were major. For 4 (16%) patients, the major difference was due to discrepancy in celiac or distant mediastinal lymph node involvement. Use of automated PET volumes changed the manual PET length in 14 patients (56%): 8 minor and 6 major. Conclusions: The use of PET/CT in treatment planning for esophageal cancer can affect target definition. Two PET-based techniques can also produce significantly different tumor volumes in a large percentage of patients. Further investigations to clarify the optimal use of PET/CT data in treatment planning are warranted. Ó 2008 Elsevier Inc. Esophageal cancer, PET/CT, treatment planning, radiation therapy.
INTRODUCTION
ability to decrease these large treatment volumes. Together these complex planning and delivery issues can affect both local efficacy of RT and treatment-related toxicity. [18F]-fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) employs radiolabeled glucose to ascertain the metabolic activity and glucose utilization of a tumor. Because FDG-PET measures a functional characteristic of a tumor, it theoretically can enable improved target definition for radiation planning. Multiple studies have already demonstrated that FDG-PET and FDG-PET/CT improves the accuracy of metastatic staging in esophageal cancer (8–15). One difficulty in using FDG-PET is the lack of standardization in interpretation of the signal produced by annihilation photons. The most common method of incorporation of PET data in gross tumor volume (GTV) definition is a ‘‘manual’’ approach in which the physician visually uses the PET data to modify the GTV. This approach, however, is highly subjective. Automated approaches have been suggested as a more objective way to delineate GTV. In one study evaluating
Radiation therapy (RT) is an integral component in the management of esophageal cancer. In combination with chemotherapy, it can be used preoperatively (1–3) or definitively (4, 5). However, significant toxicity can be associated with RT because of the large fields commonly employed to address the uncertainty of tumor extent and microscopic spread (6, 7). The radiation treatment planning process for esophageal cancer is complex. Because of the advanced presentation of esophageal cancers, large radiation fields are commonly employed. The treatment volume is further increased by generous longitudinal margins to acknowledge the high risk of submucosal spread (6, 7), as well as coverage of the regional nodal basins, including celiac nodes for distal tumors and supraclavicular nodes for more proximal cancers. These large fields, in turn, encompass significant volumes of critical normal organs including the heart and lungs. Inherent difficulties in target delineation and significant organ motion temper our
Reprint requests to: Theodore S. Hong, M.D., Department of Radiation Oncology, Massachusetts General Hospital, 100 Blossom Street, Cox LL, Boston, MA 02114. Tel: (617) 724-1159; Fax: (617) 726-3603; E-mail: [email protected]
Conflict of interest: none. Received June 8, 2008, and in revised form July 25, 2008. Accepted for publication July 31, 2008. 1612
PET/CT in treatment planning for esophageal cancer d T. S. HONG et al.
1613
Fig. 1. Definition of major difference in extent of primary tumor. This schematic shows two gross tumor volumes (GTVs) of identical length that differ by > 2 cm in superior and inferior extent. The black rectangle represents the theoretical field edge of the boost, classically defined as 2 cm from the GTV. As demonstrated, if the red GTV represented the contoured GTV and the blue GTV represented the true extent of tumor, a 2-cm margin around the red (contoured) GTV would result in a geographic miss.
target definition in lung cancer, GTV definition was performed by using manual and automated threshold techniques.16 This study demonstrated surprising discordance among the varying PET-based techniques. In our current study, we similarly evaluate the role of fused FDG-PET/CT using a manual and semiautomated technique in esophageal target definition and compared these with CT alone. METHODS AND MATERIALS Patients with esophageal cancer treated with chemoradiation from 2003 to 2006 at Brigham and Women’s Hospital were identified for this institutional review board–approved study. Only patients who underwent fused PET/CT-based treatment planning were included in this study.
Simulation procedures Patients were simulated supine with arms up on a CT scanner dedicated to radiation simulation (General Electric, Milwaukee, WI). In most patients, oral contrast was given immediately before simulation to aid in identification of a malignant stricture. Intravenous contrast was not used. Four-dimensional CT scanning (4D-CT) to ascertain organ motion was also not employed in this study.
PET/CT procedures Patients underwent fused 18FDG-PET/CT for diagnostic purposes, usually in a position similar to treatment position (supine, arms up). Patients obtained PET/CT scans per standard institutional protocol as previously reported.17 Generally, patients fasted for 4 hours before injection of FDG. All scans were acquired with a Discovery ST PET/CT scanner (General Electric). Images were obtained from the head to the proximal thighs. CT scans, obtained for attenuation correction and anatomic coregistration, were also performed without intravenous or oral contrast.
Table 1. Evaluable patients Demographics Sex M F Tumor location Cervical Upper thoracic Midthoracic Lower thoracic Gastroesophageal junction Pretreatment stage Tumor stage T1 T2 T3 T4 Not available Nodal stage N0 N1 Not available Metastasis stage M0 M1a M1b Not available AJCC stage I IIA IIB III IVA IVB Not available
No. patients (%) 21 (84%) 4 (16%) 0 (0%) 1(4%) 3 (12%) 10 (40%) 12 (48%) 0 (0%) 6 (24%) 16 (64%) 0 (0%) 3 (12%) 8 (32%) 15 (60%) 2 (8%) 21 (84%) 2 (8%) 0 (0%) 2 (8%) 0 (0%) 8 (32%) 2 (8%) 10 (40%) 2 (8%) 0 (0%) 3 (12%)
Abbreviation: AJCC = American Joint Committee on Cancer.
I. J. Radiation Oncology d Biology d Physics
1614
Volume 72, Number 5, 2008
Fig. 2. Major difference in distant ‘‘regional’’ nodal definition. The top images show a gastrohepatic node deemed involved based on CT scan (left), but in fact is PET (–) (right). In the same patient, the lower images show a subcarinal node deemed equivocal on CT (left) in fact shows [18F]-fluoro-2-deoxy-D-glucose avidity (right) and was included in the treatment field.
Fusion The PET/CT image was then fused to the treatment planning CT using the GE treatment planning system (General Electric). The fusion was performed using an automated fusion method. Generally, the physicist chose three anatomical points on the PET/CT and the planning CT. A rigid registration was then performed and was further manually adjusted if necessary. The treating physician then verified the quality of the fusion.
GTV definition 1. CT-based: A physician with no prior knowledge of each case was provided with comprehensive clinical data including history and physical examination, diagnostic contrast-enhanced CT scans, barium swallow (if available), upper endoscopy, and endoscopic ultrasound, as well as results of any biopsies of the primary tumor or suspicious nodal disease. The physician was specifically blinded to the staging PET scan. The physician did, however, know that every patient was deemed an appropriate candidate for RT, which at our institution includes patients with Stage M1a disease. 2. Manual PET/CT-based: The treating physician, as a course of standard clinical practice, had all the same clinical information but additionally had the PET/CT fused in the treatment planning system as described earlier. Typically, the PET/CT was also reviewed with nuclear medicine before contouring. The physician had the choice of contouring with the PET data overlayed with the planning CT or side by side. Most commonly, side-by-side contouring was used. The actual GTV was based on the planning CT, but the coregistered PET/CT was used to determine length of GTV and abnormal nodes.
3. Semiautomated PET/CT based: Using the PET/CT data, the GTV was defined using a method previously validated with pathologic correlation.17 This semiautomated method uses the mean activity of the liver + 2 standard deviations as a threshold value for target definition of the primary GTV. Because this definition could encompass obvious nontarget tissue, such as the heart, these automated volumes were then further edited by the nuclear medicine physician to respect anatomical boundaries clearly visible on the diagnostic CT. Because this method was previously validated only for primary tumor length, only the primary tumor was analyzed with this method.
Scoring differences in GTV Significant differences were based on the potential to significantly change radiotherapy volumes. Minor differences were scored if the superior or inferior extent (or both) of primary tumor differed by 1–2 cm (inclusive). Major differences were scored if the superior or inferior extent (or both) of primary tumor differed by > 2 cm (rather than a 2-cm difference in length) or involvement of different nodal regions were identified beyond the longitudinal extent of the primary tumor. This definition was chosen as opposed to length or volume as boost fields have historically been defined as GTV + 2 cm (1, 5). Figure 1 demonstrates schematically an example of a major length difference with two defined GTVs of identical length but different superior and inferior extent with the resultant risk of geographical miss. If the difference in the superior and inferior extent of the primary tumor was < 1 cm, it was scored no difference.
Statistical analysis Difference in mean length between CT-based GTV and manual PET-based GTV and also manual PET-based GTV and
PET/CT in treatment planning for esophageal cancer d T. S. HONG et al.
1615
Fig. 3. Major difference in gross tumor volumes (GTV) definition between CT-based and manual PET/CT-based techniques. The top images show the CT-based GTV, and the bottom images show the PET/CT based images. Note the CT-based GTV encompasses the clearly thickened portion of the esophagus (top left). However, the PET/CT-based GTV includes esophagus that appears normal on CT (bottom left) but is clearly [18F]-fluoro-2-deoxy-D-glucose avid (bottom right). semiautomated GTV were analyzed using a paired Wilcoxon test. SPSS 10.0 software (SPSS, Chicago, IL) was used for statistical analysis.
RESULTS Twenty-seven patients were identified. Two were excluded because of prior therapies. Table 1 shows patient characteristics. Most patients (84%) were male. Median age was 61 years. Of the 25 patients available for analysis, 12 had gastroesophageal junction tumors, 10 had distal esophageal tumors, 3 had midthoracic tumors, and 1 had an upper thoracic tumor. Patients by clinical and radiographic staging (including PET) had Stage IIa–IVa disease. Of the 25 evaluable patients, 23 had adenocarcinoma (92%) and 2 had squamous cell carcinoma. Most of the tumors demonstrating adenocarcinoma were moderately differentiated (8 or 23) or poorly differentiated (11 of 23). The average maximum standardized uptake value (SUV) of the primary tumor was 19 (range, 3–52). CT versus manual PET/CT GTV definition When comparing CT-based GTVs to manually defined PET/CT-based GTVs, the use of PET changed volumes for 21 of 25 (84%) patients: 12 of the changes in volume (48%) were minor, and 9 (36%) were major. Four patients (16%) had a major difference due to a difference in celiac or distant mediastinal lymph node definition. In one patient, a node separate from the primary tumor was identified as involved on PET/CT but not on CT. In another patient, a ‘‘regional’’ node distant from the primary tumor was identified as involved on CT but not PET/CT. In two
patients, CT-only and PET/CT identified different nodal groups separate from the primary tumor. An example of discrepancy in involved nodal definition is shown in Fig. 2. Seven patients (28%) had a major difference, and 12 patients (48%) had a minor difference because of the extent of the primary tumor. For three patients, the inferior border differed by > 2 cm. For three patients, the superior border differed by > 2 cm. For one patient both the superior and inferior differed by > 2 cm. Figure 3 shows an example of a major difference in superior extent of the GTV between CT-based and manual PET/CT-based definitions. Major differences are summarized in Table 2. Two of the nine patients with major difference had major differences by both nodal definition and primary tumor definition.
Table 2. CT-based GTV versus Manual PET/CT-based GTV: major differences Major difference
N (%)
Total (nodal and primary definitions) Nodal PET (+) / CT (–) CT (+) / PET (–) PET/CT discordance Primary GTV Superior border > 2 cm different PET higher CT higher Inferior border > 2 cm different PET lower CT lower
9 (36) 4 (16) 1 (4) 1 (4) 2 (8) 7 (28) 4 (16) 4 (16) 0 (0) 4 (16) 3 (12) 1 (4)
Abbreviation: GTV = gross tumor volume.
1616
I. J. Radiation Oncology d Biology d Physics
Table 3. Manual PET/CT-based GTV versus semiautomated PET/CT-based GTV: major differences (primary tumor only) Major difference
N (%)
Primary GTV Superior border > 2 cm different Manual PET higher Semiautomated PET higher Inferior border > 2 cm different Manual PET lower Semiautomated PET lower
6 (24) 3 (12) 2 (8) 1 (4) 4 (16) 3 (12) 1 (4)
Abbreviation: GTV = gross tumor volume.
PET/CT defined primary tumors were longer than CT defined in 17 of 25 (68%) of patients. Mean length was 6.53 cm and 4.80 cm for manual PET/CT-defined GTV and CT-based GTV, respectively (p = 0.008). Manual PET/CT versus semiautomated PET/CT GTV definition (primary tumor only) The semiautomated PET volumes changed the manual PET volumes in 14 patients (56%): 8 (32%) minor and 6 (24%) major. For two patients, the inferior border differed by > 2cm. For three patients, the superior border differed by > 2 cm. For one patient, the inferior and superior border both differed by > 2 cm. Major differences are summarized in Table 3. An example of a major difference is shown in Fig. 4, with the manual PET/CT-based GTV in red and the semiautomated PET/CT-based GTV in blue. The average length of GTV for the two methods of contouring did not significantly differ, with lengths of 6.53 cm and 5.73 cm for manual PET/CT- and semiautomated PET/ CT-based GTVs, respectively (p = 0.119). As with the manual PET/CT-based GTVs, the semiautomated PET/CT basedGTVs were longer than CT-based-GTVs: mean length was 5.73 cm and 4.80 cm for semiautomated PET/CT-defined GTV and CT-based GTV, respectively (p = 0.049). DISCUSSION FDG-PET/CT-based treatment planning has been enthusiastically embraced by radiation oncologists. For esophageal cancer in particular, several publications have demonstrated
Volume 72, Number 5, 2008
improved accuracy of PET-based staging of esophageal cancer (8–15). Furthermore, other publications have documented that the inclusion of PET data changes GTV definitions (10, 16–18). In this study, we evaluated two questions. First, we examined whether PET information, when added to all other available clinical data, would significantly change the identification of gross disease, or GTV definitions. Because this was a practical question regarding the impact of field design, we chose to define major differences in a way that would, at the very least, lead to a significant change in the boost volume. Because a practical definition of the boost volume used in U.S. cooperative group trials has been GTV with a 2-cm margin (superior and inferior) to block edge, it was felt that a > 2 cm change in either the superior or inferior direction would unequivocally lead to a geographic miss. In reality, a 1-cm change, when accounting for dosimetric penumbra and setup variation, would likely lead to significant underdosing of gross disease. Hence, changes of 1–2 cm were scored as minor variations for the purpose of this study. Length was not used as the primary method to score differences because length alone may miss a significant change in field borders between defined GTVs if the center of the defined GTV differed, as shown in Fig. 1. Also differences in noncontiguous gross nodal disease definitions can affect both the initial treatment volume and boost volume definition and hence were also scored as major differences. Consistent with the published literature highlighting the improvement of staging of distant nodal or metastatic disease (14), we found a difference in distant ‘‘regional’’ nodal definition in 4 of 25 patients (16%). However, we also found that PET/CT fusion changed the superior or inferior extent (or both) of the GTV by > 2 cm in 28% of patients and by 1–2 cm in another 48% of patients. It is possible that some of this difference is due to interobserver variability, because two physicians performed contouring. However, whereas interobserver variation may have accounted for some of the minor differences, the cases showing a major difference usually had a region of FDG avidity highlighting a different segment of esophagus than the region appearing thickened on CT, as seen in Fig. 3.
Fig. 4. Major difference in gross tumor volumes definition between manual PET/CT-based (red) and semiautomated PET/ CT-based (blue) definitions. The middle panel shows the PET windowing used by the physician to manually contour the GTV. However, windowing in such a way to highlight an intensity of (Ilivermean + 2 SD), the superior extent of the GTV is > 2 cm lower.
PET/CT in treatment planning for esophageal cancer d T. S. HONG et al.
The second question we evaluated was whether semiautomated PET-based target definition of the primary tumor differed significantly from the commonly used, more subjective method of target definition. The semiautomated tumor segmentation technique employed a combination of tumor maximum intensity and liver intensity to define a lower tumor intensity threshold to define the extent of tumor. This method had been previously validated at our institution and produced good correlation between PET/CT length and pathologic length (19). The semiautomated segmentation produced major differences in 24% of patients compared with the manual PET/CT based target definition. Minor differences were seen in another 32% of patients. We did not analyze the impact of the semiautomated technique for nodal differences as the technique had previously only been validated for the primary tumor. In this study, we demonstrate the difficulty of integrating PET/CT-based contouring in radiation planning. The incorporation of PET/CT into esophageal planning seems to aid in identification of distant nodal groups. However, our study shows significant differences among tumor segmentation techniques. These results are consistent with published reports that have demonstrated improved staging in distant disease but not locoregional disease (14). Our study, however, highlights one major concern in PET-based treatment planning: the lack of consensus on how best to segment tumor from normal tissue. Differences in segmentation methods can lead to significant interobserver variability. In the most qualitative approach, the planner can manually adjust the windowing of the PET images in the treatment planning system and the apparent tumor length can subjectively change. Another approach is to use a semiquantitative approach. One such method is to use SUV as a threshold level to define tumor. Two fundamental problems exist with this approach. First, the SUV concept is limited by the need for accurate recording of time of injection and scanning, as well as accurate calculation of patient size. Second, it is unknown what lower threshold should be used. One common practice for lung cancer is to use a minimal SUV threshold of 2.5 (20, 21). Alternatively, acknowledging that each tumor has a unique SUV, one can define tumor enclosed by an area defined as a percentage of maximum intensity. However, this approach can lead to drastic differences in tumor length. A final approach is to use an automated tumor segmentation method whereby a lower threshold can be calculated based on internal control (such as liver), as was done in our study (22, 23). However, no standard approach currently exists, nor has any approach been validated with pathology in a large group of patients. Other studies specifically evaluating the impact of FDGPET on esophageal treatment planning have used either the subjective manual approach (10, 16, 17) or standardization of the liver background to window the signal followed by
1617
subjective interpretation of the FDG avid area with no further windowing allowed (18). In the study that validated the method used in this study, only 17 patients had untreated tumors available for pathological evaluation (19). The resultant GTV variability created by differing utilization of PET data is highlighted by Fig. 4. The manual PET/CT-based GTV in this patient is > 2 cm higher in superior extent. In the middle panel, the displayed subjective PET windowing led the physician to include the increased superior extent manually. However, the panel on the right shows a windowing level that highlights the semiautomated threshold of 2 standard deviations greater than mean liver signal and the resulting lower superior extent of the GTV. Another limitation of PET scanning is that the acquisition of data occurs over a prolonged period of time. Inherently, PET scanning lacks reliable resolution beyond 5 mm. This uncertainty is further increased by artifacts due to internal organ motion, which can also significantly affect the determination of SUV. This limitation may be partially overcome by the simultaneous use of 4-D PET/CT scanning. Alternatively, one may view this limitation as a potential advantage, in that the organ motion captured during the acquisition of PET data may reflect organ motion during treatment. Ultimately, GTV and CTV definition must be based on clinical judgment. In our study, the primary GTV by either PET-based segmentation method tended to be longer than the GTV as determined without the aid of PET data. It is possible that incorporating PET may identify a portion of the submucosal spread that is difficult to visualize with barium swallow or CT scan, which has necessitated the traditional large elective longitudinal margin. Perhaps this elective longitudinal margin can be further decreased with integration of PET information. Further studies are clearly necessary, however, to guide how best to incorporate PET data into treatment planning.
CONCLUSIONS The use of PET/CT in treatment planning for esophageal cancer appears to provide clinically meaningful data with a significant impact on target definition when compared with CT alone. Two different PET-based techniques can also produce significantly different tumor volumes in a large percentage of patients. The automated segmentation method has the theoretical advantage of being more objective and less influenced by potentially arbitrary influences such as the window and level settings; however, because there is no gold standard for definition of the target volume, particularly in patients who receive neo-adjuvant therapy before surgery, further investigations to clarify the optimal use of PET/CT data in treatment planning are warranted.
REFERENCES 1. Tepper J, Krasna MJ, Niedzwiecki D, et al. Phase III trial of trimodality therapy with cisplatin, fluorouracil, radiotherapy, and
surgery compared with surgery alone for esophageal cancer: CALGB 9781. J Clin Oncol 2008;26:1086–1092.
1618
I. J. Radiation Oncology d Biology d Physics
2. Urba SG, Orringer MB, Turrisi A, et al. Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. J Clin Oncol 2001;19: 305–313. 3. Walsh TN, Noonan N, Hollywood D, et al. A comparison of multimodal therapy and surgery for esophageal adenocarcinoma. N Engl J Med 1996;335:462–467. 4. Herskovic A, Martz K, al-Sarraf M, et al. Combined chemotherapy and radiotherapy compared with radiotherapy alone in patients with cancer of the esophagus. N Engl J Med 1992;326: 1593–1598. 5. Minsky BD, Pajak TF, Ginsberg RJ, et al. INT 0123 (Radiation Therapy Oncology Group 94-05) Phase III trial of combinedmodality therapy for esophageal cancer: High-dose versus standard-dose radiation therapy. J Clin Oncol 2002;20:1167–1174. 6. X-s Gao, Qiao X, Wu F, et al. Pathological analysis of clinical target volume margin for radiotherapy in patients with esophageal and gastroesophageal junction carcinoma. Int J Radiat Oncol Biol Phys 2007;67:389–396. 7. Lam KY, Ma LT, Wong J. Measurement of extent of spread of oesophageal squamous carcinoma by serial sectioning. J Clin Pathol 1996;49:124–129. 8. Flamen P, Lerut A, Van Cutsem E, et al. Utility of positron emission tomography for the staging of patients with potentially operable esophageal carcinoma. J Clin Oncol 2000;18: 3202–3210. 9. Katsoulis IE, Wong WL, Mattheou AK, et al. Fluorine-18 fluorodeoxyglucose positron emission tomography in the preoperative staging of thoracic oesophageal and gastro-oesophageal junction cancer: A prospective study. Int J Surg 2007;5: 399–403. 10. Konski A, Doss M, Milestone B, et al. The integration of 18-fluoro-deoxy-glucose positron emission tomography and endoscopic ultrasound in the treatment-planning process for esophageal carcinoma. Int J Radiat Oncol Biol Phys 2005;61: 1123–1128. 11. Liberale G, Van Laethem JL, Gay F, et al. The role of PET scan in the preoperative management of oesophageal cancer [see comment]. Eur J Surg Oncol 2004;30:942–947. 12. Malik V, Keogan M, Gilham C, et al. FDG-PET scanning in the management of cancer of the oesophagus and oesophagogastric junction: Early experience with 100 consecutive cases. Irish J Med Sci 2006;175:48–54. 13. Meyers BF, Downey RJ, Decker PA, et al. The utility of positron emission tomography in staging of potentially operable carcinoma of the thoracic esophagus: Results of the American
Volume 72, Number 5, 2008
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
College of Surgeons Oncology Group Z0060 trial. J Thor Cardiovasc Surg 2007;133:738–745. Sihvo EI, Rasanen JV, Knuuti MJ, et al. Adenocarcinoma of the esophagus and the esophagogastric junction: Positron emission tomography improves staging and prediction of survival in distant but not in locoregional disease. J Gastroint Surg 2004;8: 988–996. Stahl A, Stollfuss J, Ott K, et al. FDG PET and CT in locally advanced adenocarcinomas of the distal oesophagus. Clinical relevance of a discordant PET finding. Nuclearmedizin 2005; 44:249–255. quiz N255–246. Moureau-Zabotto L, Touboul E, Lerouge D, et al. Impact of CT and 18F-deoxyglucose positron emission tomography image fusion for conformal radiotherapy in esophageal carcinoma. Int J Radiat Oncol Biol Phys 2005;63:340–345. Moureau-Zabotto L, Touboul E, Lerouge D, et al. [Impact of computed tomography (CT) and 18F-deoxyglucose positron emission tomography (FDG-PET) image fusion for conformal radiotherapy in esophageal carcinoma]. Cancer Radiotherapie 2005;9:152–160. Gondi V, Bradley K, Mehta M, et al. Impact of hybrid fluorodeoxyglucose positron-emission tomography/computed tomography on radiotherapy planning in esophageal and non–smallcell lung cancer. Int J Radiat Oncol Biol Phys 2007;67: 187–195. Mamede M, Fakhri GE, Abreu-e-Lima P, et al. Pre-operative estimation of esophageal tumor metabolic length in FDG-PET images with surgical pathology confirmation. Ann Nucl Med 2006;21:553–562. Nestle U, Kremp S, Schaefer-Schuler A, et al. Comparison of different methods for delineation of 18F-FDG PET-positive tissue for target volume definition in radiotherapy of patients with non–small cell lung cancer. J Nucl Med 2005;46:1342–1348. Nestle U, Kremp S, Grosu A-L. Practical integration of [18F]FDG-PET and PET-CT in the planning of radiotherapy for non–small cell lung cancer (NSCLC) The technical basis, ICRU-target volumes, problems, perspectives. Radiother Oncol 2006;81:209–225. Gondi V, Bradley K, Mehta M, et al. Impact of hybrid fluorodeoxyglucose positron-emission tomography/computed tomography on radiotherapy planning in esophageal and non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2007;67:187–195. Heron DE, Andrade RS, Flickinger J, et al. Hybrid PET-CT simulation for radiation treatment planning in head-and-neck cancers: A brief technical report. Int J Radiat Oncol Biol Phys 1419;60:1419–1424.
doi:10.1016/j.ijrobp.2008.07.061
PHYSICS CONTRIBUTION
IMPACT OF MANUAL AND AUTOMATED INTERPRETATION OF FUSED PET/CT DATA ON ESOPHAGEAL TARGET DEFINITIONS IN RADIATION PLANNING THEODORE S. HONG, M.D.,* JOSEPH H. KILLORAN, PH.D.,yz MARCELO MAMEDE, M.D., PH.D., y AND HARVEY J. MAMON, M.D., PH.D. * Department of Radiation Oncology, Massachusetts General Hospital, Department of yRadiation Oncology and z Nuclear Medicine, Brigham and Womens’ Hospital, Boston, MA Purpose: We compare CT-only based esophageal tumor definition with two PET/CT based methods: (1) manual contouring and (2) a semiautomated method based on specific thresholds. Methods and Materials: Patients with esophageal cancer treated at Brigham and Women’s Hospital from 2003 to 2006 were identified. CT-based tumor volumes were compared with manual PET/CT-based volumes and semiautomated PET-based tumor volumes. Differences were scored as (1) minor if the superior or inferior extent of the primary tumor (or both) differed by 1–2 cm and (2) major if the difference was > 2 cm or if different noncontiguous nodal regions were identified as being grossly involved. Results: Comparing CT-based gross tumor volumes (GTVs) to manually defined PET/CT-based GTVs, use of PET changed volumes for 21 of 25 (84%) patients: 12 patients (48%) exhibited minor differences, whereas for 9 patients (36%), the differences were major. For 4 (16%) patients, the major difference was due to discrepancy in celiac or distant mediastinal lymph node involvement. Use of automated PET volumes changed the manual PET length in 14 patients (56%): 8 minor and 6 major. Conclusions: The use of PET/CT in treatment planning for esophageal cancer can affect target definition. Two PET-based techniques can also produce significantly different tumor volumes in a large percentage of patients. Further investigations to clarify the optimal use of PET/CT data in treatment planning are warranted. Ó 2008 Elsevier Inc. Esophageal cancer, PET/CT, treatment planning, radiation therapy.
INTRODUCTION
ability to decrease these large treatment volumes. Together these complex planning and delivery issues can affect both local efficacy of RT and treatment-related toxicity. [18F]-fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) employs radiolabeled glucose to ascertain the metabolic activity and glucose utilization of a tumor. Because FDG-PET measures a functional characteristic of a tumor, it theoretically can enable improved target definition for radiation planning. Multiple studies have already demonstrated that FDG-PET and FDG-PET/CT improves the accuracy of metastatic staging in esophageal cancer (8–15). One difficulty in using FDG-PET is the lack of standardization in interpretation of the signal produced by annihilation photons. The most common method of incorporation of PET data in gross tumor volume (GTV) definition is a ‘‘manual’’ approach in which the physician visually uses the PET data to modify the GTV. This approach, however, is highly subjective. Automated approaches have been suggested as a more objective way to delineate GTV. In one study evaluating
Radiation therapy (RT) is an integral component in the management of esophageal cancer. In combination with chemotherapy, it can be used preoperatively (1–3) or definitively (4, 5). However, significant toxicity can be associated with RT because of the large fields commonly employed to address the uncertainty of tumor extent and microscopic spread (6, 7). The radiation treatment planning process for esophageal cancer is complex. Because of the advanced presentation of esophageal cancers, large radiation fields are commonly employed. The treatment volume is further increased by generous longitudinal margins to acknowledge the high risk of submucosal spread (6, 7), as well as coverage of the regional nodal basins, including celiac nodes for distal tumors and supraclavicular nodes for more proximal cancers. These large fields, in turn, encompass significant volumes of critical normal organs including the heart and lungs. Inherent difficulties in target delineation and significant organ motion temper our
Reprint requests to: Theodore S. Hong, M.D., Department of Radiation Oncology, Massachusetts General Hospital, 100 Blossom Street, Cox LL, Boston, MA 02114. Tel: (617) 724-1159; Fax: (617) 726-3603; E-mail: [email protected]
Conflict of interest: none. Received June 8, 2008, and in revised form July 25, 2008. Accepted for publication July 31, 2008. 1612
PET/CT in treatment planning for esophageal cancer d T. S. HONG et al.
1613
Fig. 1. Definition of major difference in extent of primary tumor. This schematic shows two gross tumor volumes (GTVs) of identical length that differ by > 2 cm in superior and inferior extent. The black rectangle represents the theoretical field edge of the boost, classically defined as 2 cm from the GTV. As demonstrated, if the red GTV represented the contoured GTV and the blue GTV represented the true extent of tumor, a 2-cm margin around the red (contoured) GTV would result in a geographic miss.
target definition in lung cancer, GTV definition was performed by using manual and automated threshold techniques.16 This study demonstrated surprising discordance among the varying PET-based techniques. In our current study, we similarly evaluate the role of fused FDG-PET/CT using a manual and semiautomated technique in esophageal target definition and compared these with CT alone. METHODS AND MATERIALS Patients with esophageal cancer treated with chemoradiation from 2003 to 2006 at Brigham and Women’s Hospital were identified for this institutional review board–approved study. Only patients who underwent fused PET/CT-based treatment planning were included in this study.
Simulation procedures Patients were simulated supine with arms up on a CT scanner dedicated to radiation simulation (General Electric, Milwaukee, WI). In most patients, oral contrast was given immediately before simulation to aid in identification of a malignant stricture. Intravenous contrast was not used. Four-dimensional CT scanning (4D-CT) to ascertain organ motion was also not employed in this study.
PET/CT procedures Patients underwent fused 18FDG-PET/CT for diagnostic purposes, usually in a position similar to treatment position (supine, arms up). Patients obtained PET/CT scans per standard institutional protocol as previously reported.17 Generally, patients fasted for 4 hours before injection of FDG. All scans were acquired with a Discovery ST PET/CT scanner (General Electric). Images were obtained from the head to the proximal thighs. CT scans, obtained for attenuation correction and anatomic coregistration, were also performed without intravenous or oral contrast.
Table 1. Evaluable patients Demographics Sex M F Tumor location Cervical Upper thoracic Midthoracic Lower thoracic Gastroesophageal junction Pretreatment stage Tumor stage T1 T2 T3 T4 Not available Nodal stage N0 N1 Not available Metastasis stage M0 M1a M1b Not available AJCC stage I IIA IIB III IVA IVB Not available
No. patients (%) 21 (84%) 4 (16%) 0 (0%) 1(4%) 3 (12%) 10 (40%) 12 (48%) 0 (0%) 6 (24%) 16 (64%) 0 (0%) 3 (12%) 8 (32%) 15 (60%) 2 (8%) 21 (84%) 2 (8%) 0 (0%) 2 (8%) 0 (0%) 8 (32%) 2 (8%) 10 (40%) 2 (8%) 0 (0%) 3 (12%)
Abbreviation: AJCC = American Joint Committee on Cancer.
I. J. Radiation Oncology d Biology d Physics
1614
Volume 72, Number 5, 2008
Fig. 2. Major difference in distant ‘‘regional’’ nodal definition. The top images show a gastrohepatic node deemed involved based on CT scan (left), but in fact is PET (–) (right). In the same patient, the lower images show a subcarinal node deemed equivocal on CT (left) in fact shows [18F]-fluoro-2-deoxy-D-glucose avidity (right) and was included in the treatment field.
Fusion The PET/CT image was then fused to the treatment planning CT using the GE treatment planning system (General Electric). The fusion was performed using an automated fusion method. Generally, the physicist chose three anatomical points on the PET/CT and the planning CT. A rigid registration was then performed and was further manually adjusted if necessary. The treating physician then verified the quality of the fusion.
GTV definition 1. CT-based: A physician with no prior knowledge of each case was provided with comprehensive clinical data including history and physical examination, diagnostic contrast-enhanced CT scans, barium swallow (if available), upper endoscopy, and endoscopic ultrasound, as well as results of any biopsies of the primary tumor or suspicious nodal disease. The physician was specifically blinded to the staging PET scan. The physician did, however, know that every patient was deemed an appropriate candidate for RT, which at our institution includes patients with Stage M1a disease. 2. Manual PET/CT-based: The treating physician, as a course of standard clinical practice, had all the same clinical information but additionally had the PET/CT fused in the treatment planning system as described earlier. Typically, the PET/CT was also reviewed with nuclear medicine before contouring. The physician had the choice of contouring with the PET data overlayed with the planning CT or side by side. Most commonly, side-by-side contouring was used. The actual GTV was based on the planning CT, but the coregistered PET/CT was used to determine length of GTV and abnormal nodes.
3. Semiautomated PET/CT based: Using the PET/CT data, the GTV was defined using a method previously validated with pathologic correlation.17 This semiautomated method uses the mean activity of the liver + 2 standard deviations as a threshold value for target definition of the primary GTV. Because this definition could encompass obvious nontarget tissue, such as the heart, these automated volumes were then further edited by the nuclear medicine physician to respect anatomical boundaries clearly visible on the diagnostic CT. Because this method was previously validated only for primary tumor length, only the primary tumor was analyzed with this method.
Scoring differences in GTV Significant differences were based on the potential to significantly change radiotherapy volumes. Minor differences were scored if the superior or inferior extent (or both) of primary tumor differed by 1–2 cm (inclusive). Major differences were scored if the superior or inferior extent (or both) of primary tumor differed by > 2 cm (rather than a 2-cm difference in length) or involvement of different nodal regions were identified beyond the longitudinal extent of the primary tumor. This definition was chosen as opposed to length or volume as boost fields have historically been defined as GTV + 2 cm (1, 5). Figure 1 demonstrates schematically an example of a major length difference with two defined GTVs of identical length but different superior and inferior extent with the resultant risk of geographical miss. If the difference in the superior and inferior extent of the primary tumor was < 1 cm, it was scored no difference.
Statistical analysis Difference in mean length between CT-based GTV and manual PET-based GTV and also manual PET-based GTV and
PET/CT in treatment planning for esophageal cancer d T. S. HONG et al.
1615
Fig. 3. Major difference in gross tumor volumes (GTV) definition between CT-based and manual PET/CT-based techniques. The top images show the CT-based GTV, and the bottom images show the PET/CT based images. Note the CT-based GTV encompasses the clearly thickened portion of the esophagus (top left). However, the PET/CT-based GTV includes esophagus that appears normal on CT (bottom left) but is clearly [18F]-fluoro-2-deoxy-D-glucose avid (bottom right). semiautomated GTV were analyzed using a paired Wilcoxon test. SPSS 10.0 software (SPSS, Chicago, IL) was used for statistical analysis.
RESULTS Twenty-seven patients were identified. Two were excluded because of prior therapies. Table 1 shows patient characteristics. Most patients (84%) were male. Median age was 61 years. Of the 25 patients available for analysis, 12 had gastroesophageal junction tumors, 10 had distal esophageal tumors, 3 had midthoracic tumors, and 1 had an upper thoracic tumor. Patients by clinical and radiographic staging (including PET) had Stage IIa–IVa disease. Of the 25 evaluable patients, 23 had adenocarcinoma (92%) and 2 had squamous cell carcinoma. Most of the tumors demonstrating adenocarcinoma were moderately differentiated (8 or 23) or poorly differentiated (11 of 23). The average maximum standardized uptake value (SUV) of the primary tumor was 19 (range, 3–52). CT versus manual PET/CT GTV definition When comparing CT-based GTVs to manually defined PET/CT-based GTVs, the use of PET changed volumes for 21 of 25 (84%) patients: 12 of the changes in volume (48%) were minor, and 9 (36%) were major. Four patients (16%) had a major difference due to a difference in celiac or distant mediastinal lymph node definition. In one patient, a node separate from the primary tumor was identified as involved on PET/CT but not on CT. In another patient, a ‘‘regional’’ node distant from the primary tumor was identified as involved on CT but not PET/CT. In two
patients, CT-only and PET/CT identified different nodal groups separate from the primary tumor. An example of discrepancy in involved nodal definition is shown in Fig. 2. Seven patients (28%) had a major difference, and 12 patients (48%) had a minor difference because of the extent of the primary tumor. For three patients, the inferior border differed by > 2 cm. For three patients, the superior border differed by > 2 cm. For one patient both the superior and inferior differed by > 2 cm. Figure 3 shows an example of a major difference in superior extent of the GTV between CT-based and manual PET/CT-based definitions. Major differences are summarized in Table 2. Two of the nine patients with major difference had major differences by both nodal definition and primary tumor definition.
Table 2. CT-based GTV versus Manual PET/CT-based GTV: major differences Major difference
N (%)
Total (nodal and primary definitions) Nodal PET (+) / CT (–) CT (+) / PET (–) PET/CT discordance Primary GTV Superior border > 2 cm different PET higher CT higher Inferior border > 2 cm different PET lower CT lower
9 (36) 4 (16) 1 (4) 1 (4) 2 (8) 7 (28) 4 (16) 4 (16) 0 (0) 4 (16) 3 (12) 1 (4)
Abbreviation: GTV = gross tumor volume.
1616
I. J. Radiation Oncology d Biology d Physics
Table 3. Manual PET/CT-based GTV versus semiautomated PET/CT-based GTV: major differences (primary tumor only) Major difference
N (%)
Primary GTV Superior border > 2 cm different Manual PET higher Semiautomated PET higher Inferior border > 2 cm different Manual PET lower Semiautomated PET lower
6 (24) 3 (12) 2 (8) 1 (4) 4 (16) 3 (12) 1 (4)
Abbreviation: GTV = gross tumor volume.
PET/CT defined primary tumors were longer than CT defined in 17 of 25 (68%) of patients. Mean length was 6.53 cm and 4.80 cm for manual PET/CT-defined GTV and CT-based GTV, respectively (p = 0.008). Manual PET/CT versus semiautomated PET/CT GTV definition (primary tumor only) The semiautomated PET volumes changed the manual PET volumes in 14 patients (56%): 8 (32%) minor and 6 (24%) major. For two patients, the inferior border differed by > 2cm. For three patients, the superior border differed by > 2 cm. For one patient, the inferior and superior border both differed by > 2 cm. Major differences are summarized in Table 3. An example of a major difference is shown in Fig. 4, with the manual PET/CT-based GTV in red and the semiautomated PET/CT-based GTV in blue. The average length of GTV for the two methods of contouring did not significantly differ, with lengths of 6.53 cm and 5.73 cm for manual PET/CT- and semiautomated PET/ CT-based GTVs, respectively (p = 0.119). As with the manual PET/CT-based GTVs, the semiautomated PET/CT basedGTVs were longer than CT-based-GTVs: mean length was 5.73 cm and 4.80 cm for semiautomated PET/CT-defined GTV and CT-based GTV, respectively (p = 0.049). DISCUSSION FDG-PET/CT-based treatment planning has been enthusiastically embraced by radiation oncologists. For esophageal cancer in particular, several publications have demonstrated
Volume 72, Number 5, 2008
improved accuracy of PET-based staging of esophageal cancer (8–15). Furthermore, other publications have documented that the inclusion of PET data changes GTV definitions (10, 16–18). In this study, we evaluated two questions. First, we examined whether PET information, when added to all other available clinical data, would significantly change the identification of gross disease, or GTV definitions. Because this was a practical question regarding the impact of field design, we chose to define major differences in a way that would, at the very least, lead to a significant change in the boost volume. Because a practical definition of the boost volume used in U.S. cooperative group trials has been GTV with a 2-cm margin (superior and inferior) to block edge, it was felt that a > 2 cm change in either the superior or inferior direction would unequivocally lead to a geographic miss. In reality, a 1-cm change, when accounting for dosimetric penumbra and setup variation, would likely lead to significant underdosing of gross disease. Hence, changes of 1–2 cm were scored as minor variations for the purpose of this study. Length was not used as the primary method to score differences because length alone may miss a significant change in field borders between defined GTVs if the center of the defined GTV differed, as shown in Fig. 1. Also differences in noncontiguous gross nodal disease definitions can affect both the initial treatment volume and boost volume definition and hence were also scored as major differences. Consistent with the published literature highlighting the improvement of staging of distant nodal or metastatic disease (14), we found a difference in distant ‘‘regional’’ nodal definition in 4 of 25 patients (16%). However, we also found that PET/CT fusion changed the superior or inferior extent (or both) of the GTV by > 2 cm in 28% of patients and by 1–2 cm in another 48% of patients. It is possible that some of this difference is due to interobserver variability, because two physicians performed contouring. However, whereas interobserver variation may have accounted for some of the minor differences, the cases showing a major difference usually had a region of FDG avidity highlighting a different segment of esophagus than the region appearing thickened on CT, as seen in Fig. 3.
Fig. 4. Major difference in gross tumor volumes definition between manual PET/CT-based (red) and semiautomated PET/ CT-based (blue) definitions. The middle panel shows the PET windowing used by the physician to manually contour the GTV. However, windowing in such a way to highlight an intensity of (Ilivermean + 2 SD), the superior extent of the GTV is > 2 cm lower.
PET/CT in treatment planning for esophageal cancer d T. S. HONG et al.
The second question we evaluated was whether semiautomated PET-based target definition of the primary tumor differed significantly from the commonly used, more subjective method of target definition. The semiautomated tumor segmentation technique employed a combination of tumor maximum intensity and liver intensity to define a lower tumor intensity threshold to define the extent of tumor. This method had been previously validated at our institution and produced good correlation between PET/CT length and pathologic length (19). The semiautomated segmentation produced major differences in 24% of patients compared with the manual PET/CT based target definition. Minor differences were seen in another 32% of patients. We did not analyze the impact of the semiautomated technique for nodal differences as the technique had previously only been validated for the primary tumor. In this study, we demonstrate the difficulty of integrating PET/CT-based contouring in radiation planning. The incorporation of PET/CT into esophageal planning seems to aid in identification of distant nodal groups. However, our study shows significant differences among tumor segmentation techniques. These results are consistent with published reports that have demonstrated improved staging in distant disease but not locoregional disease (14). Our study, however, highlights one major concern in PET-based treatment planning: the lack of consensus on how best to segment tumor from normal tissue. Differences in segmentation methods can lead to significant interobserver variability. In the most qualitative approach, the planner can manually adjust the windowing of the PET images in the treatment planning system and the apparent tumor length can subjectively change. Another approach is to use a semiquantitative approach. One such method is to use SUV as a threshold level to define tumor. Two fundamental problems exist with this approach. First, the SUV concept is limited by the need for accurate recording of time of injection and scanning, as well as accurate calculation of patient size. Second, it is unknown what lower threshold should be used. One common practice for lung cancer is to use a minimal SUV threshold of 2.5 (20, 21). Alternatively, acknowledging that each tumor has a unique SUV, one can define tumor enclosed by an area defined as a percentage of maximum intensity. However, this approach can lead to drastic differences in tumor length. A final approach is to use an automated tumor segmentation method whereby a lower threshold can be calculated based on internal control (such as liver), as was done in our study (22, 23). However, no standard approach currently exists, nor has any approach been validated with pathology in a large group of patients. Other studies specifically evaluating the impact of FDGPET on esophageal treatment planning have used either the subjective manual approach (10, 16, 17) or standardization of the liver background to window the signal followed by
1617
subjective interpretation of the FDG avid area with no further windowing allowed (18). In the study that validated the method used in this study, only 17 patients had untreated tumors available for pathological evaluation (19). The resultant GTV variability created by differing utilization of PET data is highlighted by Fig. 4. The manual PET/CT-based GTV in this patient is > 2 cm higher in superior extent. In the middle panel, the displayed subjective PET windowing led the physician to include the increased superior extent manually. However, the panel on the right shows a windowing level that highlights the semiautomated threshold of 2 standard deviations greater than mean liver signal and the resulting lower superior extent of the GTV. Another limitation of PET scanning is that the acquisition of data occurs over a prolonged period of time. Inherently, PET scanning lacks reliable resolution beyond 5 mm. This uncertainty is further increased by artifacts due to internal organ motion, which can also significantly affect the determination of SUV. This limitation may be partially overcome by the simultaneous use of 4-D PET/CT scanning. Alternatively, one may view this limitation as a potential advantage, in that the organ motion captured during the acquisition of PET data may reflect organ motion during treatment. Ultimately, GTV and CTV definition must be based on clinical judgment. In our study, the primary GTV by either PET-based segmentation method tended to be longer than the GTV as determined without the aid of PET data. It is possible that incorporating PET may identify a portion of the submucosal spread that is difficult to visualize with barium swallow or CT scan, which has necessitated the traditional large elective longitudinal margin. Perhaps this elective longitudinal margin can be further decreased with integration of PET information. Further studies are clearly necessary, however, to guide how best to incorporate PET data into treatment planning.
CONCLUSIONS The use of PET/CT in treatment planning for esophageal cancer appears to provide clinically meaningful data with a significant impact on target definition when compared with CT alone. Two different PET-based techniques can also produce significantly different tumor volumes in a large percentage of patients. The automated segmentation method has the theoretical advantage of being more objective and less influenced by potentially arbitrary influences such as the window and level settings; however, because there is no gold standard for definition of the target volume, particularly in patients who receive neo-adjuvant therapy before surgery, further investigations to clarify the optimal use of PET/CT data in treatment planning are warranted.
REFERENCES 1. Tepper J, Krasna MJ, Niedzwiecki D, et al. Phase III trial of trimodality therapy with cisplatin, fluorouracil, radiotherapy, and
surgery compared with surgery alone for esophageal cancer: CALGB 9781. J Clin Oncol 2008;26:1086–1092.
1618
I. J. Radiation Oncology d Biology d Physics
2. Urba SG, Orringer MB, Turrisi A, et al. Randomized trial of preoperative chemoradiation versus surgery alone in patients with locoregional esophageal carcinoma. J Clin Oncol 2001;19: 305–313. 3. Walsh TN, Noonan N, Hollywood D, et al. A comparison of multimodal therapy and surgery for esophageal adenocarcinoma. N Engl J Med 1996;335:462–467. 4. Herskovic A, Martz K, al-Sarraf M, et al. Combined chemotherapy and radiotherapy compared with radiotherapy alone in patients with cancer of the esophagus. N Engl J Med 1992;326: 1593–1598. 5. Minsky BD, Pajak TF, Ginsberg RJ, et al. INT 0123 (Radiation Therapy Oncology Group 94-05) Phase III trial of combinedmodality therapy for esophageal cancer: High-dose versus standard-dose radiation therapy. J Clin Oncol 2002;20:1167–1174. 6. X-s Gao, Qiao X, Wu F, et al. Pathological analysis of clinical target volume margin for radiotherapy in patients with esophageal and gastroesophageal junction carcinoma. Int J Radiat Oncol Biol Phys 2007;67:389–396. 7. Lam KY, Ma LT, Wong J. Measurement of extent of spread of oesophageal squamous carcinoma by serial sectioning. J Clin Pathol 1996;49:124–129. 8. Flamen P, Lerut A, Van Cutsem E, et al. Utility of positron emission tomography for the staging of patients with potentially operable esophageal carcinoma. J Clin Oncol 2000;18: 3202–3210. 9. Katsoulis IE, Wong WL, Mattheou AK, et al. Fluorine-18 fluorodeoxyglucose positron emission tomography in the preoperative staging of thoracic oesophageal and gastro-oesophageal junction cancer: A prospective study. Int J Surg 2007;5: 399–403. 10. Konski A, Doss M, Milestone B, et al. The integration of 18-fluoro-deoxy-glucose positron emission tomography and endoscopic ultrasound in the treatment-planning process for esophageal carcinoma. Int J Radiat Oncol Biol Phys 2005;61: 1123–1128. 11. Liberale G, Van Laethem JL, Gay F, et al. The role of PET scan in the preoperative management of oesophageal cancer [see comment]. Eur J Surg Oncol 2004;30:942–947. 12. Malik V, Keogan M, Gilham C, et al. FDG-PET scanning in the management of cancer of the oesophagus and oesophagogastric junction: Early experience with 100 consecutive cases. Irish J Med Sci 2006;175:48–54. 13. Meyers BF, Downey RJ, Decker PA, et al. The utility of positron emission tomography in staging of potentially operable carcinoma of the thoracic esophagus: Results of the American
Volume 72, Number 5, 2008
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
College of Surgeons Oncology Group Z0060 trial. J Thor Cardiovasc Surg 2007;133:738–745. Sihvo EI, Rasanen JV, Knuuti MJ, et al. Adenocarcinoma of the esophagus and the esophagogastric junction: Positron emission tomography improves staging and prediction of survival in distant but not in locoregional disease. J Gastroint Surg 2004;8: 988–996. Stahl A, Stollfuss J, Ott K, et al. FDG PET and CT in locally advanced adenocarcinomas of the distal oesophagus. Clinical relevance of a discordant PET finding. Nuclearmedizin 2005; 44:249–255. quiz N255–246. Moureau-Zabotto L, Touboul E, Lerouge D, et al. Impact of CT and 18F-deoxyglucose positron emission tomography image fusion for conformal radiotherapy in esophageal carcinoma. Int J Radiat Oncol Biol Phys 2005;63:340–345. Moureau-Zabotto L, Touboul E, Lerouge D, et al. [Impact of computed tomography (CT) and 18F-deoxyglucose positron emission tomography (FDG-PET) image fusion for conformal radiotherapy in esophageal carcinoma]. Cancer Radiotherapie 2005;9:152–160. Gondi V, Bradley K, Mehta M, et al. Impact of hybrid fluorodeoxyglucose positron-emission tomography/computed tomography on radiotherapy planning in esophageal and non–smallcell lung cancer. Int J Radiat Oncol Biol Phys 2007;67: 187–195. Mamede M, Fakhri GE, Abreu-e-Lima P, et al. Pre-operative estimation of esophageal tumor metabolic length in FDG-PET images with surgical pathology confirmation. Ann Nucl Med 2006;21:553–562. Nestle U, Kremp S, Schaefer-Schuler A, et al. Comparison of different methods for delineation of 18F-FDG PET-positive tissue for target volume definition in radiotherapy of patients with non–small cell lung cancer. J Nucl Med 2005;46:1342–1348. Nestle U, Kremp S, Grosu A-L. Practical integration of [18F]FDG-PET and PET-CT in the planning of radiotherapy for non–small cell lung cancer (NSCLC) The technical basis, ICRU-target volumes, problems, perspectives. Radiother Oncol 2006;81:209–225. Gondi V, Bradley K, Mehta M, et al. Impact of hybrid fluorodeoxyglucose positron-emission tomography/computed tomography on radiotherapy planning in esophageal and non-small-cell lung cancer. Int J Radiat Oncol Biol Phys 2007;67:187–195. Heron DE, Andrade RS, Flickinger J, et al. Hybrid PET-CT simulation for radiation treatment planning in head-and-neck cancers: A brief technical report. Int J Radiat Oncol Biol Phys 1419;60:1419–1424.